ALGEBRAS, DIALGEBRAS, AND POLYNOMIAL IDENTITIES MURRAYR.BREMNER 2 1 Dedicated to Yuri Bahturin on his65th birthday 0 2 n a J Abstract. This is a survey of some recent developments in the theory of 6 associative and nonassociative dialgebras, with an emphasis on polynomial 1 identitiesandmultilinearoperations. Wediscussassociative,Lie,Jordan,and alternative algebras, and the corresponding dialgebras; the KP algorithm for ] converting identities foralgebras into identities fordialgebras; the BSO algo- A rithm for converting operations in algebras into operations indialgebras; Lie andJordantriplesystems,andthecorrespondingdisystems;andanoncommu- R tativeversionofLietriplesystemsbasedonthetrilinearoperationabc−bca. . h The paper concludes witha conjecture relatingthe KP andBSO algorithms, t andsomesuggestionsforfurtherresearch. Mostoftheoriginalresultsarejoint a workwithRau´lFelipe,LuizA.Peresi,andJuanaS´anchez-Ortega. m [ 1. Algebras 1 ThroughoutthistalkthebasefieldFwillbearbitrary,butweusuallyexcludelow v characteristics, especially p ≤ n where n is the degree of the polynomial identities 9 underconsideration. Theassumptionp>nallowsustoassumethatallpolynomial 7 3 identities are multilinear and that the group algebra FSn is semisimple. 3 Definition 1.1. An algebra is a vector space A with a bilinear operation . 1 µ: A×A→A. 0 2 Unless otherwise specified, we write ab = µ(a,b) for a,b ∈ A. We say that A is 1 associative if it satisfies the polynomial identity : v (ab)c≡a(bc). i X Throughoutthispaperwewillusethesymbol≡toindicateanequationthatholds r for all values of the arguments; in this case, all a,b,c∈A. a Theorem 1.2. The free unital associative algebra on a set X of generators has basis consisting of all words of degree n≥0, x=x x ···x , where x ,x ,...,x ∈X, 1 2 n 1 2 n withtheproduct definedon basis elements byconcatenation andextendedbilinearly, (x x ···x )(y y ···y )=x x ···x y y ···y . 1 2 m 1 2 n 1 2 m 1 2 n 2010 Mathematics Subject Classification. Primary 17A30. Secondary 16R10, 17-08, 17A32, 17A40,17A50,17B60,17C05,17D05,17D10. Key words and phrases. Algebras, triple systems, dialgebras, triple disystems, polynomial identities,multilinearoperations,computer algebra. This paper is an expanded version of the lecture notes from the author’s talk at the Second International Workshop on Polynomial Identities, 2–6 September 2011, which took place at the AtlanticAlgebraCentre,MemorialUniversity,St.John’s,Newfoundland,Canada. 1 2 MURRAYR.BREMNER Definition 1.3. The commutator in an algebra is the bilinear operation [a,b]=ab−ba. This operation is anticommutative: it satisfies [a,b]+[b,a]≡0. Lemma 1.4. In an associative algebra, the commutator satisfies the identity [[a,b],c]+[[b,c],a]+[[c,a],b]≡0 (Jacobi) Definition 1.5. A Lie algebra is an algebra which satisfies anticommutativity and the Jacobi identity. Theorem 1.6. Poincar´e-Birkhoff-Witt. Every Lie algebra L has a universal associativeenvelopingalgebraU(L)forwhichthecanonicalmapL→U(L)isinjec- tive. It follows that every polynomial identity satisfied by the commutator in every associative algebra is a consequence of anticommutativity and the Jacobi identity. Remark 1.7. Most texts on Lie algebras include a proof of the PBW Theorem. The most beautiful proof is that of Bergman [2] using noncommutative Gr¨obner bases; see also de Graaf [18, Ch. 6]. For the history of the PBW Theorem, see Grivel[23]. For asurveyonGr¨obner-Shirshovbases,see BokutandKolesnikov[4]. Definition 1.8. The anticommutator in an algebra is the bilinear operation a◦b=ab+ba; we omit the scalar 1. This operationis commutative: it satisfies a◦b−b◦a≡0. 2 Lemma 1.9. In an associative algebra, the anticommutator satisfies the identity ((a◦a)◦b)◦a−(a◦a)◦(b◦a)≡0 (Jordan) Definition 1.10. A Jordan algebra is an algebra which satisfies commutativity and the Jordan identity. Theorem 1.11. There exist polynomial identities satisfied by the anticommutator in every associative algebra which do not follow from commutativity and the Jordan identity. The lowest degree in which such identities exist is 8. Remark 1.12. A Jordanalgebra is called special if it is isomorphic to a subspace ofanassociativealgebraclosedunderthe anticommutator. Anpolynomialidentity for Jordan algebras is called special if it is satisfied by all special Jordan algebras butnotbyallJordanalgebras. The firstspecialidentities forJordanalgebraswere foundbyGlennie[20,21]. Foracomputationalapproach,seeHentzel[26]. Another s-identitywasobtainedbyThedy[48];seealsoMcCrimmon[38]and[40,Appendix B.5]. For a survey on identities in Jordan algebras,see McCrimmon [39]. Remark 1.13. From the perspective of polynomial identities, there is a clear di- chotomy between the two bilinear operations, commutator and anticommutator. Both operations satisfy simple identities in low degree; for the commutator, these identitiesimplyallthe identitiessatisfiedbytheoperation,butfortheanticommu- tator, there exist special identities of higher degree. ALGEBRAS, DIALGEBRAS, AND POLYNOMIAL IDENTITIES 3 2. Dialgebras Wenowrecalltheconceptofadialgebra: avectorspacewithtwomultiplications. AssociativedialgebraswereoriginallydefinedbyLodayinthe1990s,andtheresults quotedinthissectionwereprovedbyhim;seeespeciallyhisoriginalpaper[33]and hissurveyarticle[34]. AssociativedialgebrasprovidethenaturalsettingforLeibniz algebras,a “non-anticommutative” generalization of Lie algebras; see Loday [32]. Definition 2.1. A dialgebra is a vector space A with two bilinear operations, ⊣: A×A→A, ⊢: A×A→A, called the left and right products. We say that A is a 0-dialgebra if it satisfies the left and right bar identities, (a⊣b)⊢c≡(a⊢b)⊢c, a⊣(b⊣c)≡a⊣(b⊢c). An associative dialgebra is a 0-dialgebra satisfying left, right, and inner as- sociativity: (a⊣b)⊣c≡a⊣(b⊣c), (a⊢b)⊢c≡a⊢(b⊢c), (a⊢b)⊣c≡a⊢(b⊣c). Definition 2.2. Let x = x x ···x be a monomial in an associative dialgebra, 1 2 n with some placement of parentheses and choice of operations. The center of x, denoted c(x), is defined by induction on n: • If n=1 then x=x and c(x)=x . 1 1 • If n ≥ 2 then x = y ⊣ z or x = y ⊢ z, and c(x) = c(y) or c(x) = c(z) respectively. Lemma 2.3. Let x = x x ···x be a monomial in an associative dialgebra with 1 2 n c(x) = x . Then the following expression does not depend on the placement of i parentheses: x=x ⊢···⊢x ⊢x ⊣x ⊣···⊣x . 1 i−1 i i+1 n Definition 2.4. The expression in the Lemma 2.3 is called the normal form of the monomial x, and is abbreviated using the hat notation: x=x ···x ···x . 1 i n Theorem 2.5. The free associative dialgebra on a set X of generators has basis b consisting of all monomials in normal form: x=x ···x ···x (1≤i≤n, x ,x ,...,x ∈X). 1 i n 1 2 n Two such monomials are equal if and only if they have the same permutation of b the generators and the same position of the center. The left and right products are defined on monomials as follows and extended bilinearly: x⊣y =(x ···x ···x )⊣(y ···y ···y )=x ···x ···x y ···y , 1 i n 1 j p 1 i n 1 p x⊢y =(x ···x ···x )⊢(y ···y ···y )=x ···x y ···y ···y . 1 bi n 1 bj p 1 bn 1 j p Definition 2.6. The dicommutator in a dialgebra is the bilinear operation b b b ha,bi=a⊣b−b⊢a. In general, this operation is not anticommutative. Lemma 2.7. In an associative dialgebra, the dicommutator satisfies the identity hha,bi,ci≡hha,ci,bi+ha,hb,cii (Leibniz) 4 MURRAYR.BREMNER Definition 2.8. ALeibniz algebra(orLiedialgebra)isanalgebrasatisfyingthe Leibniz identity. Remark 2.9. If we set b=c in the Leibniz identity then we obtain ha,hb,bii≡0, and the linearized form of this identity (assuming characteristic not 2) is ha,hb,cii+ha,hc,bii≡0 (right anticommutativity) Theorem 2.10. Loday-Pirashvili. Every Leibniz algebra L has a universal as- sociative enveloping dialgebra U(L) for which the canonical map L → U(L) is injective. Hence every polynomial identity satisfied by the dicommutator in every associative dialgebra is a consequence of the Leibniz identity. Remark 2.11. The Loday-Pirashvili Theorem is the generalization to dialgebras of the PBW Theorem. For the original proof, see [35]. For different approaches, see Aymon and Grivel [1], Insua and Ladra [28]. Remark2.12. Thedefinitionofassociativedialgebracanbemotivatedintermsof theLeibnizidentity. IfweexpandtheLeibnizidentityinanonassociativedialgebra using the dicommutator as the operation, then we obtain (a⊣b−b⊢a)⊣c−c⊢(a⊣b−b⊢a)≡ (a⊣c−c⊢a)⊣b−b⊢(a⊣c−c⊢a)+a⊣(b⊣c−c⊢b)−(b⊣c−c⊢b)⊢a. Equating terms with the same permutation of a,b,c gives the following identities: (a⊣b)⊣c≡a⊣(b⊣c), 0≡(a⊣c)⊣b−a⊣(c⊢b), (b⊢a)⊣c≡b⊢(a⊣c), 0≡b⊢(c⊢a)−(b⊣c)⊢a, c⊢(a⊣b)≡(c⊢a)⊣b, c⊢(b⊢a)≡(c⊢b)⊢a. These are equivalent to the identities defining associative dialgebras. Definition 2.13. Theantidicommutatorinadialgebraisthebilinearoperation a⋆b=a⊣b+b⊢a. In general, this operation is not commutative. Lemma 2.14. In an associative dialgebra, the antidicommutator satisfies a⋆(b⋆c)≡a⋆(c⋆b) (right commutativity) (b⋆a2)⋆a≡(b⋆a)⋆a2 (right Jordan identity) ha,b,c2i≡2ha⋆c,b,ci (right Osborn identity) where a2 =a⋆a and ha,b,ci=(a⋆b)⋆c−a⋆(b⋆c). Remark 2.15. These identities wereobtained independently by different authors: Vel´asquez and Felipe [49], Kolesnikov [29], Bremner [5]. A generalization of the TKK construction from Lie and Jordan algebras to Lie and Jordan dialgebras has been given by Gubarev and Kolesnikov [24]. For further work on the structure of Jordan dialgebras, see Felipe [19]. I have named the last identity in Lemma 2.14 after Osborn [41]; it is a noncommutative version of the identity stating that a commutator of multiplications is a derivation. Definition 2.16. A Jordan dialgebra (or quasi-Jordan algebra) is an algebra satisfying right commutativity and the right Jordan and Osborn identities. ALGEBRAS, DIALGEBRAS, AND POLYNOMIAL IDENTITIES 5 Remark 2.17. Strictly speaking, Leibniz algebras and Jordan dialgebras have two operations, but they are opposite, so we consider only one. This will become clear when we discuss the KP algorithm for converting identities for algebras into identities for dialgebras. Theorem 2.18. There exist special identities for Jordan dialgebras; that is, poly- nomial identities satisfied by the antidicommutator in every associative dialgebra which are not consequences of right commutativity and the right Jordan and Os- born identities. Remark 2.19. This result was obtained using computer algebra by Bremner and Peresi [11]. The lowest degree for such identities is 8; some but not all of these identities are noncommutative versions of the Glennie identity. For a theoretical approach to similar results, including generalizations of the classical theorems of Cohn, Macdonald, and Shirshov, see Voronin [50]. 3. From algebras to dialgebras We now discuss a general approachto the following problem. Problem 3.1. Given a polynomial identity for algebras, how do we obtain the corresponding polynomial identity (or identities) for dialgebras? An algorithm has been developed by Kolesnikov and Pozhidaev for converting multilinear identities for algebras into multilinear identities for dialgebras. For binary algebras, see [29]; for the generalization to n-ary algebras, see [45]. The underlying structure from the theory of operads is discussed by Chapoton [16]. Kolesnikov-Pozhidaev(KP) algorithm. Theinputisamultilinearpolynomial identity ofdegreed forann-aryoperationdenotedby the symbol{−,··· ,−}with n arguments. The output of Part 1 is a collection of d multilinear identities of degree d for n new n-ary operations denoted {−,··· ,−} for 1 ≤ i ≤ n. The i outputofPart2isacollectionofmultilinearidentitiesofdegree2n−1forthesame new operations. Part 1. Givenamultilinearidentityofdegreedinthen-aryoperation{−,··· ,−}, we describe the application of the algorithm to one monomial, and extend this by linearity to the entire identity. Let a a ...a be a multilinear monomial of degree 1 2 d dwithsomeplacementofn-aryoperationsymbols. Foreachi=1,...,dweconvert themonomiala a ...a intheoriginaln-aryoperationintoanewmonomialofthe 1 2 d same degree in the n new n-ary operations, according to the following rule, based on the position of the variable a , called the central variable of the monomial. For i each occurrence of the original n-ary operation in the monomial, either a occurs i in one of the n arguments or not, and we have two cases: (a) If a occurs in the j-th argument then we convert {−,··· ,−} to the j-th i new operation symbol {−,··· ,−} . j (b) If a does not occur in any of the n arguments, then either i • a occurs to the left of{−,··· ,−}: we convert{−,··· ,−}to the first i new operation symbol {−,··· ,−} , or 1 • a occurs to the right of {−,··· ,−}: we convert {−,··· ,−} to the i last new operation symbol {−,··· ,−} . n 6 MURRAYR.BREMNER Part 2. We also include the following identities, generalizing the bar identities for associative dialgebras, for all i,j =1,...,n with i6=j and all k,ℓ=1,...,n: {a ,...,a ,{b ,··· ,b } ,a ,...,a } ≡ 1 i−1 1 n k i+1 n j {a ,...,a ,{b ,··· ,b } ,a ,...,a } . 1 i−1 1 n ℓ i+1 n j This identity says that the n new operations are interchangeable in the i-th argu- ment of the j-th new operation when i6=j. Example 3.2. Thedefinitionofassociativedialgebracanbe obtainedbyapplying the KP algorithm to the associativity identity, which we write in the form {{a,b},c}≡{a,{b,c}}. Theoperation{−,−}producestwonewoperations{−,−} ,{−,−} . Part1ofthe 1 2 algorithm produces three identities by making a,b,c in turn the central variable: {{a,b} ,c} ≡{a,{b,c} } , {{a,b} ,c} ≡{a,{b,c} } , 1 1 1 1 2 1 1 2 {{a,b} ,c} ≡{a,{b,c} } . 2 2 2 2 Part 2 of the algorithm produces two identities: {a,{b,c} } ≡{a,{b,c} } , {{a,b} ,c} ≡{{a,b} ,c} . 1 1 2 1 1 2 2 2 Ifwewritea⊣bfor{a,b} anda⊢bfor{a,b} thenthesearethethreeassociativity 1 2 identities and the two bar identities. Example 3.3. The definition of Leibniz algebra can be obtained by applying the KP algorithm to the identities defining Lie algebras: anticommutativity (in its bilinear form) and the Jacobi identity, [a,b]+[b,a]≡0, [[a,b],c]+[[b,c],a]+[[c,a],b]≡0. Part 1 of the algorithm produces five identities: [a,b] +[b,a] ≡0, [[a,b] ,c] +[[b,c] ,a] +[[c,a] ,b] ≡0, 1 2 1 1 2 2 2 1 [a,b] +[b,a] ≡0, [[a,b] ,c] +[[b,c] ,a] +[[c,a] ,b] ≡0, 2 1 2 1 1 1 2 2 [[a,b] ,c] +[[b,c] ,a] +[[c,a] ,b] ≡0. 2 2 2 1 1 1 The two identities of degree 2 are equivalent to [a,b] ≡ −[b,a] , so the second 2 1 operationissuperfluous. Eliminatingthesecondoperationfromthethreeidentities of degree 3 shows that each of them is equivalent to the identity [[a,b] ,c] +[a,[c,b] ] −[[a,c] ,b] ≡0. 1 1 1 1 1 1 If we write ha,bi= [a,b] then we obtain a form of the Leibniz identity. Part 2 of 1 the algorithm produces two identities: [a,[b,c] ] ≡[a,[b,c] ] , [[a,b] ,c] ≡[[a,b] ,c] . 1 1 2 1 1 2 2 2 Eliminating the second operation gives right anticommutativity: ha,hb,cii+ha,hc,bii≡0. However,as we have alreadyseen in Remark 2.9, the Leibniz identity implies right anticommutativity, so it suffices to retain only the Leibniz identity. ALGEBRAS, DIALGEBRAS, AND POLYNOMIAL IDENTITIES 7 Example 3.4. To apply the KP algorithm to the defining identities for Jordan algebras, we write commutativity and the multilinear form of the Jordan identity using the operation symbol {−,−}: {a,b}−{b,a}≡0, {{{a,c},b},d}+{{{a,d},b},c}+{{{c,d},b},a} −{{a,c},{b,d}}−{{a,d},{b,c}}−{{c,d},{b,a}}≡0. From commutativity, Part 1 of the algorithm gives two identities of degree 2: {a,b} −{b,a} ≡0, {a,b} −{b,a} ≡0, 1 2 2 1 Thesetwo identities areequivalentto {a,b} ≡{b,a} : the secondoperationis the 2 1 oppositeofthe first,andso wemay eliminate {−,−} . Fromthe linearizedJordan 2 identity, Part 1 of the algorithm gives four identities of degree 4: {{{a,c} ,b} ,d} +{{{a,d} ,b} ,c} +{{{c,d} ,b} ,a} 1 1 1 1 1 1 2 2 2 −{{a,c} ,{b,d} } −{{a,d} ,{b,c} } −{{c,d} ,{b,a} } ≡0, 1 1 1 1 1 1 2 2 2 {{{a,c} ,b} ,d} +{{{a,d} ,b} ,c} +{{{c,d} ,b} ,a} 2 2 1 2 2 1 2 2 1 −{{a,c} ,{b,d} } −{{a,d} ,{b,c} } −{{c,d} ,{b,a} } ≡0, 2 1 2 2 1 2 2 1 2 {{{a,c} ,b} ,d} +{{{a,d} ,b} ,c} +{{{c,d} ,b} ,a} 2 1 1 2 2 2 1 1 1 −{{a,c} ,{b,d} } −{{a,d} ,{b,c} } −{{c,d} ,{b,a} } ≡0, 2 1 1 2 2 2 1 1 1 {{{a,c} ,b} ,d} +{{{a,d} ,b} ,c} +{{{c,d} ,b} ,a} 2 2 2 2 1 1 2 1 1 −{{a,c} ,{b,d} } −{{a,d} ,{b,c} } −{{c,d} ,{b,a} } ≡0. 2 2 2 2 1 1 2 1 1 We replace every instance of {−,−} by the opposite of {−,−} : 2 1 {{{a,c} ,b} ,d} +{{{a,d} ,b} ,c} +{a,{b,{d,c} } } 1 1 1 1 1 1 1 1 1 −{{a,c} ,{b,d} } −{{a,d} ,{b,c} } −{{a,b} ,{d,c} } ≡0, 1 1 1 1 1 1 1 1 1 {{b,{c,a} } ,d} +{{b,{d,a} } ,c} +{{b,{d,c} } ,a} 1 1 1 1 1 1 1 1 1 −{{b,d} ,{c,a} } −{{b,c} ,{d,a} } −{{b,a} ,{d,c} } ≡0, 1 1 1 1 1 1 1 1 1 {{{c,a} ,b} ,d} +{c,{b,{d,a} } } +{{{c,d} ,b} ,a} 1 1 1 1 1 1 1 1 1 −{{c,a} ,{b,d} } −{{c,b} ,{d,a} } −{{c,d} ,{b,a} } ≡0, 1 1 1 1 1 1 1 1 1 {d,{b,{c,a} } } +{{{d,a} ,b} ,c} +{{{d,c} ,b} ,a} 1 1 1 1 1 1 1 1 1 −{{d,b} ,{c,a} } −{{d,a} ,{b,c} } −{{d,c} ,{b,a} } ≡0. 1 1 1 1 1 1 1 1 1 We simplify the notation and write {a,b} as ab. The last four identities become: 1 ((ac)b)d+((ad)b)c+a(b(dc))−(ac)(bd)−(ad)(bc)−(ab)(dc)≡0, (b(ca))d+(b(da))c+(b(dc))a−(bd)(ca)−(bc)(da)−(ba)(dc)≡0, ((ca)b)d+c(b(da))+((cd)b)a−(ca)(bd)−(cb)(da)−(cd)(ba)≡0, d(b(ca))+((da)b)c+((dc)b)a−(db)(ca)−(da)(bc)−(dc)(ba)≡0. Thefirstisequivalenttothethirdandtothe fourth,soweretainonlythe firstand second. Part 2 of the algorithm produces two identities: {a,{b,c} } ≡{a,{b,c} } , {{a,b} ,c} ≡{{a,b} ,c} . 1 1 2 1 1 2 2 2 Rewriting these using only the first operation gives {a,{b,c} } ≡{a,{c,b} } , {c,{a,b} } ≡{c,{b,a} } . 1 1 1 1 1 1 1 1 8 MURRAYR.BREMNER These two identities are equivalent to right commutativity: a(bc) ≡ a(cb). We rearrangethe two retained identities of degree 4 and apply right commutativity: ((ac)b)d−(ac)(bd)+((ad)b)c−(ad)(bc)−(ab)(cd)+a(b(cd))≡0, (b(ac))d+(b(ad))c+(b(cd))a−(bd)(ac)−(bc)(ad)−(ba)(cd)≡0. The first identity can be reformulated in terms of associators as follows, (ac,b,d)+(ad,b,c)−(a,b,cd)≡0, and assuming characteristic 6=2 this is equivalent to (a,b,c2)≡2(ac,b,c). Setting a=c=d in the second identity and assuming characteristic 6=3 gives (ba2)a≡(ba)a2, Thus we obtain right commutativity and the right Osborn and Jordan identities. Example 3.5. The multilinear forms of the left and right alternative identities defining alternative algebras are: (a,b,c)+(b,a,c)≡0, (a,b,c)+(a,c,b)≡0. Expanding the associators gives (ab)c−a(bc)+(ba)c−b(ac)≡0, (ab)c−a(bc)+(ac)b−a(cb)≡0. We apply the KP algorithm to these identities, writing {−,−} for the original bi- linearoperation. Part1givessixidentitiesrelatingthetwonewoperations{−,−} 1 and {−,−} : in each of the two original identities we make either a, b, or c the 2 centralargument. Inthis case,weretainbothoperations,since thereis noidentity ofdegree2relating{−,−} and{−,−} . Weobtainsixidentitiesdefiningalterna- 1 2 tive dialgebras; in the first (second) group of three, the only differences are in the subscripts 1 and 2 indicating the position of the central variable: {{a,b} ,c} −{a,{b,c} } +{{b,a} ,c} −{b,{a,c} } ≡0, 1 1 1 1 2 1 1 2 {{a,b} ,c} −{a,{b,c} } +{{b,a} ,c} −{b,{a,c} } ≡0, 2 1 1 2 1 1 1 1 {{a,b} ,c} −{a,{b,c} } +{{b,a} ,c} −{b,{a,c} } ≡0, 2 2 2 2 2 2 2 2 {{a,b} ,c} −{a,{b,c} } +{{a,c} ,b} −{a,{c,b} } ≡0, 1 1 1 1 1 1 1 1 {{a,b} ,c} −{a,{b,c} } +{{a,c} ,b} −{a,{c,b} } ≡0, 2 1 1 2 2 2 2 2 {{a,b} ,c} −{a,{b,c} } +{{a,c} ,b} −{a,{c,b} } ≡0. 2 2 2 2 2 1 1 2 We revert to standard notation: ⊣ for {−,−} and ⊢ for {−,−} : 1 2 (a⊣b)⊣c−a⊣(b⊣c)+(b⊢a)⊣c−b⊢(a⊣c)≡0, (a⊢b)⊣c−a⊢(b⊣c)+(b⊣a)⊣c−b⊣(a⊣c)≡0, (a⊢b)⊢c−a⊢(b⊢c)+(b⊢a)⊢c−b⊢(a⊢c)≡0, (a⊣b)⊣c−a⊣(b⊣c)+(a⊣c)⊣b−a⊣(c⊣b)≡0, (a⊢b)⊣c−a⊢(b⊣c)+(a⊢c)⊢b−a⊢(c⊢b)≡0, (a⊢b)⊢c−a⊢(b⊢c)+(a⊢c)⊣b−a⊢(c⊣b)≡0. We rewrite these in terms of the left, right and inner associators: (a,b,c) +(b,a,c) ≡0, (a,b,c) +(b,a,c) ≡0, ⊣ × × ⊣ ALGEBRAS, DIALGEBRAS, AND POLYNOMIAL IDENTITIES 9 (a,b,c) +(b,a,c) ≡0, (a,b,c) +(a,c,b) ≡0, ⊢ ⊢ ⊣ ⊣ (a,b,c) +(a,c,b) ≡0, (a,b,c) +(a,c,b) ≡0. × ⊢ ⊢ × These six identities show how the associators change under various transpositions ofthearguments. Inparticular,theidentitiesinthesecondrowshowthattheright operation a ⊢ b is left alternative, and the left operation a ⊣ b is right alternative. (We do not have two alternative operations.) Part 2 of the algorithm simply gives the left and right bar identities. To summarize, we define an alternative dialgebra to be a 0-dialgebra satisfying (a,b,c) +(c,b,a) ≡0, (a,b,c) −(b,c,a) ≡0, (a,b,c) +(a,c,b) ≡0, ⊣ ⊢ ⊣ ⊢ × ⊢ where the left, right, and inner associators are defined by (a,b,c) =(a⊣b)⊣c−a⊣(b⊣c), (a,b,c) =(a⊢b)⊢c−a⊢(b⊢c), ⊣ ⊢ (a,b,c) =(a⊢b)⊣c−a⊢(b⊣c). × This definition was originally obtained in a different way by Liu [31]. Example 3.6. Malcev algebras [44] can be defined by the polynomial identities of degree ≤ 4 satisfied by the commutator in every alternative algebra. Bremner, PeresiandSa´nchez-Ortega[12]usedcomputeralgebratostudythe identitiessatis- fiedbythedicommutatorineveryalternativedialgebra,andprovedthateverysuch identityofdegree≤6isaconsequenceoftheidentitiesofdegree≤4. Theyshowed that the identities of degree ≤ 4 are equivalent to those obtained by applying the KPalgorithmtolinearizedformsofanticommutativitytheMalcevidentity,namely right anticommutativity and a “noncommutative” version of the Malcev identity: a(bc)+a(cb)≡0, ((ab)c)d−((ad)b)c−(a(cd))b−(ac)(bd)−a((bc)d)≡0. These two identities define the variety of Malcev dialgebras. 4. Multilinear operations Wenowconsidergeneralizationsofthecommutatorab−baandanticommutator ab+ba to operations of arbitrary “arity” (number of arguments). The following definitions and examples are based primarily on Bremner and Peresi [10]. Definition 4.1. A multilinear n-ary operation ω(a ,a ,...,a ), or more con- 1 2 n ciselyann-linearoperation,isalinearcombinationofpermutationsofthemonomial a a ···a regarded as an element of the free associative algebra on n generators: 1 2 n ω(a ,a ,...,a )= x a a ···a (x ∈F). 1 2 n σ σ(1) σ(2) σ(n) σ σX∈Sn We identify ω(a ,a ,...,a ) with an element of FS , the group algebra of the 1 2 n n symmetric group S which acts by permuting the subscripts of the generators. n Definition 4.2. Two multilinear operations are equivalent if each is a linear combination of permutations of the other; this is the same as saying that the two operations generate the same left ideal in FS . n Example 4.3. For n = 2, we have the Wedderburn decomposition FS ≈ F⊕F, 2 where the two simple ideals correspond to partitions 2 and 1+1 and have bases ab+ba and ab−ba respectively (writing a,b instead of a ,a ). There are four 1 2 equivalence classes, corresponding to the commutator, the anticommutator, the zero operation, and the original associative operation ab. 10 MURRAYR.BREMNER reduced matrix form permutation form 0 1 1 0, , 0 abc−bac−cab+cba (cid:20) (cid:20) 0 0 (cid:21) (cid:21) 1 1/2 2 0, , 0 abc+acb−bca−cba (cid:20) (cid:20) 0 0 (cid:21) (cid:21) 0 1 3 1, , 0 abc+cba (cid:20) (cid:20) 0 0 (cid:21) (cid:21) 1 0 4 1, , 0 abc+bac (cid:20) (cid:20) 0 0 (cid:21) (cid:21) 1 1 5 1, , 0 abc+acb (cid:20) (cid:20) 0 0 (cid:21) (cid:21) 1 1/2 6 1, , 0 2abc+acb+2bac+bca (cid:20) (cid:20) 0 0 (cid:21) (cid:21) 0 1 7 0, , 1 2abc−acb−2bac+bca (cid:20) (cid:20) 0 0 (cid:21) (cid:21) 1 −1 8 0, , 1 abc−acb (cid:20) (cid:20) 0 0 (cid:21) (cid:21) 1 2 9 0, , 1 abc−bac (cid:20) (cid:20) 0 0 (cid:21) (cid:21) 1 1/2 10 0, , 1 abc−cba (cid:20) (cid:20) 0 0 (cid:21) (cid:21) 0 1 11 1, , 1 abc−bac+bca (cid:20) (cid:20) 0 0 (cid:21) (cid:21) 1 0 12 1, , 1 abc+cab−cba (cid:20) (cid:20) 0 0 (cid:21) (cid:21) 1 1 13 1, , 1 abc+bca−cba (cid:20) (cid:20) 0 0 (cid:21) (cid:21) 1 −1 14 1, , 1 abc+bac+cab (cid:20) (cid:20) 0 0 (cid:21) (cid:21) 1 2 15 1, , 1 abc+acb+bca (cid:20) (cid:20) 0 0 (cid:21) (cid:21) 1 1/2 16 1, , 1 abc+acb+bac (cid:20) (cid:20) 0 0 (cid:21) (cid:21) 1 0 17 0, , 0 abc−bca (cid:20) (cid:20) 0 1 (cid:21) (cid:21) 1 0 18 1, , 0 abc+acb+bac−cba (cid:20) (cid:20) 0 1 (cid:21) (cid:21) 1 0 19 0, , 1 abc+acb−bca−cab (cid:20) (cid:20) 0 1 (cid:21) (cid:21) Table 1. Simplified trilinear operations